問題文
A cluster is sized so that every node is allocated when all planned jobs run. What is the consequence of that choice?
選択肢
- A single node failure leaves no node to replace it, so the affected job cannot restart until the node is repaired or another job is stopped.
- The scheduler will automatically shrink the running jobs to free a node, which makes spare capacity unnecessary, and a job simply continues with the participants that remain.
- The cluster becomes unable to accept new jobs, which is the only consequence and is resolved as soon as any job finishes, so the queue drains again the moment any running job ends.
- Nothing of consequence, on the grounds that a node that fails is removed from the pool and the job continues on the remaining nodes at a proportionally reduced speed until the node is returned to service by the maintenance team.